bugfix: correct node rank

[cauyxy]

What does this PR do?

Corrected the logic of calculating rank from dividing by num_nodes to num_gpu_per_meachine

Fixes #19436

Before submitting

• Was this discussed/agreed via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:

Reviewer checklist

• [ yes ] Is this pull request ready for review? (if not, please submit in draft mode)
• [ yes ] Check that all items from Before submitting are resolved
• [ yes ] Make sure the title is self-explanatory and the description concisely explains the PR
• [ yes ] Add labels and milestones (and optionally projects) to the PR so it can be classified

---

📚 Documentation preview 📚: https://pytorch-lightning--19437.org.readthedocs.build/en/19437/

[cauyxy]

I think this is a common problem that almost everyone who uses multiple machines and multiple cards for training will encounter. Should a quick fix be released?

[tchaton]

Hey @cauyxy, great catch !

Would you mind adding a test ?

Best,
T.C

[cauyxy]

Hey @cauyxy, great catch !

Would you mind adding a test ?

Best, T.C

Just like the previous one, it might take me some time if I were to do it myself. Can you help me out?😊

[cauyxy]

i have added test for dist training when gpu_per_node > node_num

[tchaton]

Thanks @cauyxy. This is perfect ! This test fails on master and the fix is correct ! Thanks for your contribution !